A Machine Learning Approach Applied to Energy Prediction in Job Shop Environments

Author(s)

Pereira, Moises S. ; Lima, Fabio ; IEEE

• Document type: Journal article
• Source: IECON 2018 - 44TH ANNUAL CONFERENCE OF THE IEEE INDUSTRIAL ELECTRONICS SOCIETY; v. N/A, p. 6-pg., 2018-01-01.

Abstract

Energy efficiency has become a great challenge for manufacturing companies. Although it is possible to improve efficiency applying new and more efficient machines, decision makers tend to look for some less expensive alternatives. In this context, the adoption of more efficient strategies during the production planning can allow the reduction in energy consumption and associated emissions. Furthermore, the current reality of manufacturing companies, brought by Industry 4.0 concepts, requires more flexibility of production systems, thus, increasing complexity for machine rescheduling without compromising sustainable requirements. In this paper, we propose a method to predict total energy consumption in job shop systems applying machine learning techniques. Different schedules may result in different consumption rates. However, there is a nonlinear relationship between these targets. Therefore, an Artificial Neural Network (ANN) is applied for a quick estimation of total energy consumption. In order to validate the model, computational experiments, using digital manufacturing software tools, are performed on different job shop configurations to show the efficiency of the proposed model. (AU)

• FAPESP's process: 17/25987-3 - Management of electrical energy in advanced manufacturing systems using machine learning
• Grantee: Fábio Lima
• Support Opportunities: Regular Research Grants